The Institute of Electrical and Electronics Engineers, which establishes network standards, defines a local area network (LAN) as a data communications system that enables a number of independent devices to communicate with each other in a limited geographic area. In other words, a LAN is a network of computers linked together via cable within a limited area. LANs are proprietary systems limited to a finite number of users. Consequently, LANs are not connected with public telephone and cable systems. They also generally serve an area of less than one mile, and are usually confined to a single building. Nevertheless, a computer network spanning a university campus or a large industrial site with multiple buildings also can be classified as a LAN. LANs also permit workers—isolated in separate offices—to operate off the same system, as if they were all sitting around a single computer. In addition, they allow audio, video, and data communication and have higher bandwidth connections than traditional wide area networks Besides linking computers, a LAN also may include one or more servers as well as several printers and other computer equipment, depending on the number of network users.
One of the great attributes of a LAN is that it may be installed simply, upgraded or expanded with little difficulty, and moved or rearranged without disruption. Perhaps most importantly, anyone initiated in the use of a personal computer can be trained to communicate or perform work over a LAN. Moreover, LANs can improve productivity by enabling workers to share information and databases and can save companies money by allowing them to purchase fewer computer peripherals such as printers and plotters, which can be shared via a LAN.
Local area networks have their genesis in distributed computing systems that were introduced during the 1960s. Initially, they consisted of "dumb" terminals connected to a single mainframe processor via a wiring system. Instructions entered on the terminal keyboard were registered at the mainframe, where they were processed, and a visual representation of these instructions was sent back to the terminal for display on a screen.
The first protocols for these distributed computer workstations were proprietary, meaning that they were designed specifically for equipment designed by a certain company. As a result, IBM equipment could not be mixed with Digital Equipment Corp. (DEC), Xerox Corp., Wang, or any other manufacturer's machinery.
By the late 1970s, however, several companies proposed "open" standards under which equipment from one manufacturer could be made to emulate the operating system of another. This allowed manufacturers to compete for business in systems that had previously been closed to all but the system designer.
The first of these standards was Ethernet, a "listen-and-transmit" protocol developed by DEC, Intel Corp., and Xerox. Ethernet identified the type of instructions being generated and, if necessary, conditioned them so they could be read by the mainframe on a common terminal, or bus.
In 1980 the Institute of Electrical and Electronics Engineers determined that continued reluctance to open protocols would seriously retard the growth of distributed computer systems. It established a group called the 802 committee to establish networking standards for the entire industry. This would ensure that customers could migrate from one vendor to another without sacrificing their considerable investments in existing systems. The standards also compelled manufacturers to follow the standards, or risk being dealt out of the market.
Ethernet was offered to the 802 committee as a standard. Several manufacturers balked because Ethernet did not work well under heavy traffic. Instead, the Ethernet standard was adapted into three versions, corresponding to network designs. Still, these standards could support 1,024 workstations over an end-to-end distance of two kilometers.
Processor manufacturing technology, however, had progressed so far that entire computers could be condensed into a single desktop unit. The first of these "personal computers" (PCs) was introduced by IBM in 1981.
IBM's PC featured an open bus architecture, meaning that IBM provided design specifications to other manufacturers in the hope that they would design compatible equipment and software for the system. IBM could impose this architecture on the market because it had very high market penetration and was the leading manufacturer in the industry.
The PC changed the type of information sent over office computer networks. Terminals were no longer "dumb," but contained the power to perform their own instructions and maintain their own memories. This took considerable pressure off the mainframe device, whose energies could now be devoted to more complex tasks.
An analysis of common office tasks revealed that as much as 80 percent of the work performed by the average employee never left the room in which it was produced. This factor established demand characteristics for individual, worker-specific PCs.
The remaining 20 percent of office tasks required transmission of data for access by other workers. LANs enabled this data to be directed to a common printer, serving a dozen or more workers. This eliminated the need for each worker to have a printer and ensured that the one printer provided was not underutilized.
In addition, LANs allowed data to be called up directly on other workers' computers, providing immediate communication and eliminating the need for paper. The most common application was in interoffice communications, or electronic mail (e-mail). Messages could be directed to one or several people and copied to several more over the LAN.
As a result, an e-mail system became something of an official record of communications between workers. Addressees became obligated to respond to e-mail messages in a timely manner because their failure to answer could be documented for supervisors.
PCs transformed LANs from mere shared processors to fully integrated communication devices. In fact, developments in processing technology endowed some PCs with even greater capacity than the mainframe computers to which they were attached. For some applications, the need for a mainframe was completely eliminated. With processing power distributed among PCs, the mainframe's main role was eclipsed. While still useful for complex processing, administrative functions and data file storage became the job of a new device, the file server.
A local area network generally requires three principal components besides the computers being connected: network cards, cable or wire, and software. While software-driven, the physical properties of a LAN include interfaces, called network access units, which connect computers to networks. These units are actually network cards installed on computer motherboards. Their job is to provide a connection, monitor availability of access, set or buffer the data transmission speed, ensure against transmission errors and collisions, and assemble data from the LAN into usable form for the PC. A LAN consisting of two to four computers, however, can be created without a network card and this kind of network is called a slotless system. In a slotless system, the computers' serial and parallel ports are connected to each other. Such LANs are very inexpensive and businesses use them largely for sharing hard-drive space and printers, but they cannot support high-speed data transmission.
The next part of a LAN is the wiring, which provides the physical connection from one PC to another, and to servers and printers and other peripherals. The properties of the wiring determine transmission speeds.
The first LANs were connected with coaxial cable, a variety of the type used to deliver cable television. Certain kinds of coaxial cable are relatively inexpensive and coaxial cable is simple to attach. More importantly, these cables provide great bandwidth (the system's rate of data transfer), enabling transmission speeds up to 20 megabits per second.
During the 1980s, however, AT&T introduced a LAN wiring system using ordinary twisted wire pair of the type used for telephones. The primary advantages of twisted wire pair are that it is very cheap, simpler to splice than coaxial, and is already installed in many buildings as obsolete or redundant wiring. In fact, many buildings were left with stranded 25-pair wiring once used for key telephone systems.
But the downside of this simplicity is that its bandwidth is more limited, meaning that twisted pair, designed for voice communication, transmits data at a slow rate. For example, AT&T's first LAN product, StarLAN, had a capacity of only one megabit per second. Subsequent improvements expanded this capacity tenfold and eliminated the need for shielded, or conditioned, wiring.
A more recent development in LAN wiring is fiber distributed data interface (FDDI) or fiber-optic cable. This type of wiring uses thin strands of glass to transmit pulses of light between terminals. Its advantages are that it provides tremendous bandwidth and thus allows very high transmission speeds: data transmission at a rate of up to 100 megabits per second. And, because it is optical rather than electronic, it is impervious to electromagnetic interference. Fiber optics also supports a network of up to 1,000 computers and can transmit signals up to 50 miles. Its main drawbacks, however, are that splicing is difficult and requires a high degree of skill and that it costs far more than its counterparts.
The primary application of fiber is not between terminals, but between LAN buses (terminals) located on different floors. As a result, FDDI is used mainly in building risers. Within individual floors, LAN facilities remain coaxial or twisted wire pair.
Where a physical connection cannot be made, such as across a street or between buildings where easements for wiring cannot be secured, microwave radio may be used. It is often difficult, however, to secure frequencies for this medium.
Another alternative in this application is light transceivers, which project a beam of light similar to fiber-optic cable, but through the air, rather than over cable. These systems do not have the frequency allocation or radiation problems associated with microwave, but they are susceptible to interference from fog and other obstructions.
The software needed for a LAN depends on the kind of network being created: whether slotless, peer-to-peer, or server-based. Slotless system software usually enables users to perform the rudimentary tasks associated with slotless systems: sharing hard-drive space and printers. In addition, this software may provide e-mail and security features. Basic slotless system utilities are included in standard operating systems such as Windows 95 and Windows 98.
Peer-to-peer software facilitates peer-to-peer networks, which allow all users to access and use the resources of all computers attached to the network, including hard drives and printers. Peer-to-peer LANs, however, generally lack security and administration capabilities of server-based LANs, as well as the capacity for large data transmissions. Nevertheless, they are less expensive than their server-based counterparts.
Client/server software is designed for networks that designate a computer as the hub or server of the LAN. The server computer is linked to all the other computers of the network—the client computers—and it carries out the majority of the server duties, such as user access control and coordinating user tasks. The server cannot be used as a workstation, however, without special software allowing it to function as one. The client computers use the programs and data stored on the server. Server-based networks are best for companies with heavy network traffic.
LANs are designed in several different topologies or physical patterns of connecting terminals. The most common topology is the bus, where several terminals are connected directly to each other over a single transmission path. Its layout is linear and it resembles a street with several driveways. The bus network requires cables that allow signals to flow in either direction, called a full duplex medium. Each terminal on the bus LAN contends with other terminals for access to the system. When it has secured access to the system, it broadcasts its message to all the terminals at once. The message is picked up by the one terminal or group of terminal stations for which it is intended. The bus network's lack of routing and central control make it very reliable, because failure of one of the network's computers generally will not impede the flow of other network traffic.
A second topology, the star network, also works like a bus in terms of contention and broadcast. But in the star, stations are connected to a single, central node that administers access. The central node knows the path to all the other nodes, which makes routing easy. The central node also enables access control and establishing a priority status for users. Several of these nodes may be connected to one another. For example, a bus serving 6 stations may be connected to another bus serving 10 stations and a third bus connecting 12 stations. The star topology is most often used where the connecting facilities are coaxial or twisted wire pair.
The ring topology connects each station to its own node, and these nodes are connected in a circular fashion. Node I is connected to node 2, which is connected to node 3, and so on, and the final node is connected back to node 1. Messages sent over the LAN are regenerated by each node, but retained only by the addressees. Eventually, the message circulates back to the sending node, which removes it from the stream. Consequently, this configuration does not require routing.
LANs are effective because their transmission capacity is greater than any single terminal on the system. As a result, each station terminal can be offered a certain amount of time on the LAN, like a time-sharing arrangement. To take advantage of this window of opportunity, stations organize their messages into compact packets that can be quickly disseminated.
In contending for access, a station with something to send stores its data packet in a buffer until the LAN is clear. At that point the message is sent out. Sometimes, two stations may detect the opening at the same time and send their messages simultaneously. Unaware that another message has been sent out, the two signals will collide on the LAN. When this happens it is up to the software to determine who should go first and ask both machines to try again.
In busy LANs, collisions would occur all the time, slowing the system down considerably. To solve the problem, the LAN software circulates a token. This works like a ticket that is distributed only to one station at a time. Instead of waiting for the LAN to clear, the station waits to receive the token.
When it has the token, the station sends its packet out over the LAN. When it is done, it returns the token to the stream for the next user. Tokens, used in ring and bus topologies, virtually eliminate the problem of collisions by providing orderly, noncontention access.
The transmission methods used on LANs are either baseband or broadband. The baseband medium uses a high-speed digital signal consisting of square wave DC voltage. While it is fast, it can accommodate only one message at a time. As a result it is suitable for smaller networks where contention is low. It also is very simple, requiring no tuning or frequency discretion circuits. As a result, the transmission medium may be connected directly to the network access unit and is suitable for use over twisted wire pair facilities.
In contrast, the broadband medium tunes signals to special frequencies, much like cable television. Stations are instructed by signaling information to tune to a specific channel to receive information. The information within each channel on a broadband medium may also be digital, but they are separated from other messages by frequency. As a result, the medium generally requires higher capacity cables, such as coaxial cable. Suited for busier LANs, broadband systems require the use of tuning devices in the network access unit that can filter out all but the single channel it needs.
File and printer servers provided the initial impetus for companies to develop LANs so that they could share databases and expensive peripherals such as printers. Furthermore, the heart of the LAN, the administrative software, generally resides either in a dedicated file server (which functions as a server only) or, in a smaller, less busy LAN, in a computer acting as a file server (which also can function as a workstation). In addition to acting as a kind of traffic cop by controlling and regulating user access, this server holds files for shared use in its hard drives, administers applications such as operating systems, and coordinates tasks such as printing.
Where a single computer is used both as a workstation and a file server, response times may lag because its processors are forced to perform several instructions at once. In addition, the system will store certain files on different computers connected with the LAN. Consequently, if one machine is down, the entire system may be crippled. Moreover, if the system were to crash due to undercapacity, some data may be lost or corrupted.
The addition of a dedicated file server may be costly, but it provides several advantages over a distributed system. In addition to ensuring access even when some machines are down, it is unencumbered by multiple duties. Its only jobs are to hold files and provide access.
Since 1990 one of the most notable developments in LANs has been the growth of communication servers that allow LANs to communicate with networks outside of the LANs themselves. Communication servers enable remote LAN access, e-mail, fax, and other communication services. Like other servers, this one controls access and facilitates use of communications software and hardware. As with file and printer servers, a separate computer may be designated as a dedicated server to enhance reliability.
Furthermore, the LANs of the late 1990s began to include servers devoted other applications such as those for decision support, transaction processing, and data warehousing. The number of application servers is forecast to increase significantly as more companies add dedicated application servers to their LANs.
The speed of the LAN is measured in terms of throughput, a figure different from transmission speed because it takes into account the capacity of the wiring and the distance between stations. The data rate, which most directly represents response time, is determined by throughput and other factors such as overhead bits and other signals, error and collision recovery, software and hardware efficiency, and the memory capacity of disk drives.
As mentioned earlier, LANs are generally limited in size because of the physical properties of the network: distance, impedance (a kind of electrical resistance), and load. Some equipment, such as repeaters, can extend the range of a LAN. Repeaters have no processing ability, but simply regenerate signals that are weakened by impedance.
Other types of LAN equipment with processing ability include gateways, which refer to the hardware and software necessary to enable technologically different networks to communicate with each other. A gateway, for example, can compensate for dissimilar protocols to pass information by translating them into a simpler code, such as ASCII. A bridge works like a gateway, but instead of connecting technologically different networks, it connects networks employing the same kind of technology. Similar to a bridge, a router is the hardware and software connection between two (or more) networks or subnetworks that routes traffic from one network or subnetwork to another. But routers primarily control the transmission of packets to their destinations.
Gateways, bridges, and routers can act as repeaters, boosting signals over greater distances. They also enable separate LANs located in different buildings to communicate with each other.
In some cases, separate LANs located in different cities—and even separate countries—may be linked over a public network. Whether these are "nailed up" dedicated links or switched services, the connection of two or more such LANs in separate geographic locations is referred to as a wide area network (WAN).
WANs require the use of special software programs in the operating system to enable dial-up connections that may be performed by a router. Unless limited to modem speeds, these connections may require special services, such as integrated services digital network (ISDN), to ensure efficient transmission, particularly of large data files. Increasingly, companies employing LANs in separate locations also operate WANs.
Another device, which can be used to create LANs, is the private branch exchange. Private branch exchanges (PBXs) are telephone switching systems that generally serve one company or network and route data and information to specific servers, rather than broadcasting to all stations. PBXs are oblivious to operating systems and use only twisted pair. As a result, PBX networks are somewhat slower and their applications are more limited than other kinds of LANs.
LANs are susceptible to many kinds of transmission errors. Electromagnetic interference from motors, power lines, and sources of static, as well as shorts from corrosion, can corrupt data. In addition, different kinds of cables are more susceptible to these problems than others. Software bugs and hardware failures can also introduce errors, as can irregularities in wiring and connections.
LANs generally compensate for these errors by working off an uninterruptible power source, such as batteries, and using backup software to recall most recent activity and hold unsaved material. Some systems may be designed for redundancy, such as keeping two file servers and alternate wiring to route around failures.
In addition, as computer software evolves requiring faster processors and faster rates of transmission, LAN technology also must evolve. Multimedia and video applications in particular force companies to upgrade their LANs in order to use such applications in a network environment. Consequently, LANs increasingly need to transmit data at gigabit, not megabit, speeds and hence older technology must be upgraded or replaced.
When purchasing a LAN, or even investigating the possibility of installing one, several considerations must be kept in mind. The costs involved and the administrative support needed often far exceed reasonable predictions.
Three general concerns when considering a LAN include administration, security, and productivity. Administration utilities regulate and coordinate file, application, peripheral, and resource use, while security utilities control access to the network. Productivity refers to the tasks a company wants to perform via a LAN, which may include file, database, and printer sharing. Moreover, thorough consideration of potential costs should include such factors as purchase price of equipment, spare parts and taxes, installation costs, labor and building modifications, and permits. Operating costs include forecasted public network traffic, diagnostics, and routine maintenance. In addition, the buyer should seek a schedule of potential costs associated with upgrades and expansion of the network, since company LANs tend to require new technology and to expand periodically.
The vendor should agree to a contract expressly detailing the degree of support that will be provided in installing and turning on the system. In addition, the vendor should provide a maintenance contract that binds the company to make immediate, free repairs when performance of the system exceeds prescribed standards. All of these factors should be addressed in the buyer's request for proposal, which is distributed to potential vendors.
[ John Simley ,
updated by Karl Heil ]
Daines, Bernard. "The Future of Gigabit LANs." Telecommunications, January 1997, 15.
Derrick, Dan. Network Know-How: Concepts, Cards, and Cables. Berkeley, CA: Osborne/McGraw-Hill, 1992.
Green, Harry James. The Business One Irwin Handbook of Telecommunications. 2nd ed. Homewood, IL: Business One Irwin, 1991.
Madron, Thomas W. Local Area Networks. New York: John Wiley & Sons, Inc., 1994.
Rhodes, Peter D. Building a Network. New York: McGraw-Hill, 1996.